EECE 301 Signals & Systems Prof.

Mark Fowler
Note Set #14
C-T Signals: Fourier Transform (for Non-Periodic Signals) Reading Assignment: Section 3.4 & 3.5 of Kamen and Heck

Course Flow Diagram

The arrows here show conceptual flow between ideas. Note the parallel structure between the pink blocks (C-T Freq. Analysis) and the blue blocks (D-T Freq. Analysis). New Signal Models
Ch. 1 Intro
C-T Signal Model Functions on Real Line System Properties LTI Causal Etc D-T Signal Model Functions on Integers

Ch. 3: CT Fourier Signal Models

Fourier Series Periodic Signals Fourier Transform (CTFT) Non-Periodic Signals

Ch. 5: CT Fourier System Models

Frequency Response Based on Fourier Transform

Ch. 6 & 8: Laplace Models for CT Signals & Systems

Transfer Function

New System Model

New System Model

Ch. 2 Diff Eqs

C-T System Model Differential Equations D-T Signal Model Difference Equations Zero-State Response Zero-Input Response Characteristic Eq.

Ch. 2 Convolution
C-T System Model Convolution Integral D-T Signal Model Convolution Sum

New System Model

New Signal Model Powerful Analysis Tool

Ch. 4: DT Fourier Signal Models

DTFT (for Hand Analysis) DFT & FFT (for Computer Analysis)

Ch. 5: DT Fourier System Models

Freq. Response for DT Based on DTFT

Ch. 7: Z Trans. Models for DT Signals & Systems

Transfer Function

New System Model

New System Model

4.3 Fourier Transform

Recall: Fourier Series represents a periodic signal as a sum of sinusoids or complex sinusoids

e jk0t

Note: Because the FS uses harmonically related frequencies k0, it can only create periodic signals

Q: Can we modify the FS idea to handle non-periodic signals? A: Yes!! What about x(t ) =
c k e j k t ?

k =

That will give some non-periodic signals but not some that are important!! The problem with x(t ) = ck e jk t is that it cannot include all possible k = frequencies!

How about:

1 x(t ) = 2

X ( )e jt d
Plays the role of ck

Yes this will work for any practical non-periodic signal!!

Plays the role of jk0t

Called the Fourier Integral also, more commonly, called the Inverse Fourier Transform

Integral replaces sum because it can add up over the continuum of frequencies!

Okay given x(t) how do we get X()?

X ( ) = x(t )e jt dt

Note: X() is complex-valued function of (-, ) |X()|

X ( )
Need to use two plots to show it

Comparison of FT and FS Fourier Series: Used for periodic signals Fourier Transform: Used for non-periodic signals (although we will see later that it can also be used for periodic signals) Synthesis Fourier Series Analysis

x(t ) =

n =

c e
k

jk0t

1 t 0 +T ck = x(t )e jk0t dt T t0
Fourier Coefficients

Fourier Series

Fourier 1 x(t ) = 2 Transform

X ( )e

jk0t

X ( ) = x(t )e jt dt

Inverse Fourier Transform

Fourier Transform

FS coefficients ck are a complex-valued function of integer k FT X() is a complex-valued function of the variable (-, )

Synthesis Viewpoints:
FS:

x(t ) =

n =

ck e jk0t

|ck| shows how much there is of the signal at frequency k0

ck shows how much phase shift is needed at frequency k0

FT:

1 x(t ) = 2

X ( )e jt d

|X()| shows how much there is in the signal at frequency

X() shows how much phase shift is needed at frequency

Some FT Notation:
If X() is the Fourier transform of x(t) then we can write this in several ways: 1. x (t ) X ( ) 2. X ( ) = F {x (t )} 3. x (t ) = F
1

F{ } is an operator that operates on x(t) to give X()

{X ( )}

F-1{ } is an operator that operates on X() to give x(t)

Analogy: Looking at X() is like looking at an x-ray of the signal- in the sense that an x-ray lets you see what is inside the object shows what stuff it is made from. In this sense: X() shows what is inside the signal it shows how much of each complex sinusoid is inside the signal Note: x(t) completely determines X() X() completely determines x(t) There are some advanced mathematical issues that can be hurled at these comments well not worry about them

FT Example: Decaying Exponential

Given a signal x(t) = e-btu(t) find X() if b > 0

Solution: First see what x(t) looks like:

x(t )
b controls decay rate

t
The u(t) part forces this to zero

Nowapply the definition of the Fourier transform. Recall the general form:

X ( ) = x (t )e jt dt

Now plug in for our signal:

X ( ) = e u (t )e
bt

jt

dt = e e
0

bt jt

dt = e (b + j ) t dt
0

integrand = 0 for t < 0 due to the u(t)

Set lower limit to 0 and then u(t) = 1 over integration range

Easy integral!

1 ( b+ j ) t 1 e = e ( b+ j ) e ( b+ j ) 0 = b + j t =0 b + j

t =

1 1 b j 0 [0 1] = e e 2 e = { 13 { b + j b + j =0 mag =1 =1 1 = b + j

Summary of FT Result for Decaying Exponential

x (t ) = e u (t )
For b > 0

bt

1 X ( ) = b + j
(Complex Valued)

X ( ) =

1 b +
2
1

Magnitude

X ( ) = tan Phase b
x (t ) = e bt u(t ) 1 b > 0 controls
decay rate

X ( )

Fourier Transform of e-btu(t) for b = 10

(volts)

True for every real-valued signal

MATLAB Commands to Compute FT w=-100:0.2:100; b=10; X=1./(b+j*w); Note: Books Fig. 3.12 only shows one-sided spectrum plots Plotting Commands subplot(2,1,1); plot(w,abs(X)) xlabel('Frequency \omega (rad/sec)') ylabel('|X(\omega|) (volts)'); grid subplot(2,1,2); plot(w,angle(X)) xlabel('Frequency \omega (rad/sec)') ylabel('<X(\omega) (rad)'); grid

Time Signal
|X( )|

10

Fourier Transform
b=0.1

x(t)

b=0.1
0.5

0 -10 1

10 t (sec)

20

30

40

0 -100 1

-50

0 (rad/sec)

50

100

Exploring Effect of decay rate b on the Fourier Transforms Shape

0.5

|X( )|

b=1
x(t)

b=1
0.5

0 -10 1

10 t (sec)

20

30

40

0 -100 0.1

-50

0 (rad/sec)

50

100

0.5

|X( )|

b=10
x(t)

b=10
0.05

0 -10

10 t (sec)

20

30

40

0 -100

-50

0 (rad/sec)

50

100

Note: As b increases 1. Decay rate in time signal increases 2. High frequencies in Fourier transform are more prominent.

Example: FT of a Rectangular pulse

p (t )

= pulse width

Given: a rectangular pulse signal p(t) Find: P() the FT of p(t)

Note the Notational Convention: lower-case for time signal and corresponding upper-case for its FT

Solution: Note that

1, t 2 2 p (t ) = 0, otherwise

Now apply the definition of the FT:

P ( ) = p (t )e jt dt =

/2

/ 2

e jt dt

Limit integral to where p(t) is nonzero and use the fact that it is 1 over that region

1 jt = e j

2 2

j j 2 e 2 e 2 = j2

Artificially inserted 2 in numerator and denominator Use Eulers Formula

= sin 2

2 sin 2 P ( ) =

1/ decays down as || gets big this causes the overall function to decay down

For this case the FT is real valued so we can plot it using a single plot (shown in solid blue here):

= 1/2
2/ 2/

-2/

-2/

2 sin 2 P ( ) =

The sin wiggles up down between 2/

Now lets think about how to make magnitude/phase plot

Even though this FT is real-valued we can still plot it using magnitude and phase plots: We can view any real number as a complex number that has zero as its imaginary part

Im R Re

A positive real number R will have: |R| = R R = 0

Im + R - Re

A negative real number R will have: |R| = -R R =

Applying these Ideas to the Real-valued FT P()

Phase = 0 Phase =

Here I have chosen - to display odd symmetry

Effect of Pulse Width on the FT P()

1 |X( )| x(t) 0.5 0 -4 1 |X( )| x(t) 2

=2
-3 -2 -1 0 1 t (sec) 2 3 4

=2
-50 0 (rad/sec) 50 100

0 -100 1

0.5 0 -4 1

=1
-3 -2 -1 0 1 t (sec) 2 3 4

0.5

=1
-50 0 (rad/sec) 50 100

0 -100 1 |X( )|

x(t)

0.5 0 -4

= 1/2
-3 -2 -1 0 1 t (sec) 2 3 4

0.5

= 1/2
-50 0 (rad/sec) 50 100

0 -100

Note: As width decreases, FT is more widely spread

Definition of Sinc Function

The result we just found had this mathematical form: This kind of structure shows up frequently enough that we define a special function to capture it:

2 sin 2 P ( ) =

sin(x ) Define: sinc ( x ) = x

Plot of sinc(x)
Note that sinc(0) = 0/0. So Why is sinc(0) = 1?

With a little manipulation we can re-write the FT result for a pulse in terms of the sinc function:
Recall:

sinc ( x ) =

sin(x ) x

Need times something Now we need the same thing down here as inside the sine

2 sin 2 sin 2 sin 2 2 2 P ( ) = = =

2 sin sin 2 = sinc 2 = = 2 2 2 2

P ( ) = sinc 2

Table of Common Fourier Transform Results

We have just found the FT for two common signals
x (t) = e
1, p ( t ) = 0,
bt

u (t)

X ( ) =

1 b + j

otherwise

P ( ) = sinc 2

There are tables in the book but I recommend that you use the Tables I provide on the Website

See FT Table on the Course Website for a list of these and many other FT. You should study this table If you encounter a time signal or FT that is on this table you should recognize that it is on the table without being told that it is there. You should be able to recognize entries in graphical form as well as in equation form (so it would be a good idea to make plots of each function in the table to learn what they look like! See next slide!!!) You should be able to use multiple entries together with the FT properties well learn in the next set of notes (and there will be another Table!)

For your FT Table you should spend time making sketches of the entries like this:

X ()

t
P ()

p (t)

Bandlimited and Timelimited Signals

Now that we have the FT as a tool to analyze signals, we can use it to identify certain characteristics that many practical signals have. A signal x(t) is timelimited (or of finite duration) if there are 2 numbers T1 & T2 such that:

x(t ) = 0 t [T1 , T2 ]

T1

T2

A (real-valued) signal x(t) is bandlimited if there is a number B such that

X ( ) = 0
X ( )
2B 2B

> 2B

Recall: If x(t) is real-valued then |X()| has even symmetry

FACT: A signal can not be both timelimited and bandlimited

Any timelimited signal is not bandlimited Any bandlimited signal is not timelimited

Note: All practical signals must start & stop

timelimited Practical signals are not bandlimited!

But engineers say practical signals are effectively bandlimited because for almost all practical signals |X()| decays to zero as gets large
X ( )
2B 2B

FT of pulse

Some application-specific level that specifies small enough to be negligible

This signal is effectively bandlimited to B Hz because |X()| falls below (and stays below) the specified level for all above 2B

Bandwidth (Effective Bandwidth)

Abbreviate Bandwidth as BW

For a lot of signals like audio they fill up the lower frequencies but then decay as gets large: X ( )

2B 2B

We say the signals BW = B in Hz if there is negligible content for || > 2B

Must specify what negligible means

For Example: 1. 2. High-Fidelity Audio signals have an accepted BW of about 20 kHz A speech signal on a phone line has a BW of about 4 kHz
Early telephone engineers determined that limiting speech to a BW of 4kHz still allowed listeners to understand the speech

For other kinds of signals like radio frequency (RF) signals they are concentrated at high frequencies
X ( )

1 = 2f1

2 = 2f 2

If the signals FT has negligible content for || [1, 2] then we say the signals BW = f2 - f1 in Hz For Example: 1. The signal transmitted by an FM station has a BW of 200 kHz = 0.2 MHz a. The station at 90.5 MHz on the FM Dial must ensure that its signal does not extend outside the range [90.4, 90.6] MHz b. Note that: FM stations all have an odd digit after the decimal point. This ensures that adjacent bands dont overlap: i. FM90.5 covers [90.4, 90.6] ii. FM90.7 covers [90.6, 90.8], etc. 2. The signal transmitted by an AM station has a BW of 20 kHz a. A station at 1640 kHz must keep its signal in [1630, 1650] kHz b. AM stations have an even digit in the tens place and a zero in the ones